RAMS

Digital Twin-based Prognostics and Health Management for Subsea

systems: Concepts, Opportunities and Challenges

PhD Candidate : Mohsin Abbas

Supervisor : Prof. Yiliu Liu

PhD Project Title : Condition Based Maintenance decisions and Digital Twin
for subsea systems.

Date: 04.03.2021
 Motivation & Objectives

converge existing perspective for better

understanding of DT and its difference to
traditional simulation approaches.

Outline opportunities and challenges that

DT based PHM brings with an aim of
defining future research direction.

2
 Outline

• Part I: DT concepts and difference between DT approach and

traditional simulation approach.

• Part II: DT types and classification.

• Part III : Subsea maintenance

• Part IV: DT based PHM : Opportunites and Challenges

3
Part I: DT concepts and differences between DT approach and traditional
simulation approach.
 History of DT
• Dates back to 2002 when Grieves introduced a
conceptual model in his course for product life cycle
management.

• Preliminary model included three parts: physical

product, digital replica, and their connections.

• Firstly, named as Mirrored Space Model (MSM) in

2005, Information Mirroring model in 2006 and was
later revisited by NASA in 2012 calling it as a Digital
Twin

5
 Digital Twin Definitions
First Definition of Digital Twin by NASA:

Glaessgen and Stargel [10] “A Digital Twin is an integrated Multiphysics, multiscale, probabilistic simulation of an as-built vehicle or system
that uses the best available physical models, sensor updates, fleet history, etc., to mirror the life of its
corresponding flying twin”.

In Production:
Maurer [17], Tao et al. [12]
“the DT is a real digital mapping depicting the production process and product performance of all components in
the product life cycle using physical data, virtual data and interaction data between them”.

In Engineering:
Bajaj et al. [29], Reifsnider & Majumdar [26]
“A unified Ultra-high-fidelity system model that can coordinate architecture, mechanical, electrical, software,
verification, and other discipline- specific models across the system lifecycle, federating models in multiple
vendor tools and configuration-controlled repositories.”

In terms of the purpose of the digital twin:

Macchi et al. [27], Bolton et al. [13], Gabor et al.
“The DT can be considered as a dynamic virtual entity, relying on the sensed and transmitted data from the
[16], Soderbergh et al. [14]
existing system, with the purpose to; enable understanding, learning and reasoning, and allow for more accurate
simulation in different scales of time and space for optimizations, control and decision-making.”

In aviation:
Tuegel [28]
“Ultra-realistic, cradle-to-grave computer model of an aircraft structure that is used to assess the aircraft’s ability
to meet mission requirements.

In industry player terms:

Siemens [36]
“digital copy that is created and developed simultaneously with the real machine”

In Health:
Ulrik [38] “transformative platform for patient specific hypertensive intervention”

6
In General

A DT is a real time :

- virtual replica of a physical entity

- entailing various models

- to control physical counter part, provide

services and assist in decision making.

7
 DT Solution Architecture

8
 Difference Between DT approach and Traditional simulation approach

▪ Real Time - The DT approach utilizes real-time data to study its physical counterpart
behavior, while simulations are mostly based on historical data, assumed, or randomly
generated data.
¨ - The time lapse between the synchronization of DT and its physical counterpart may
also be utilized as one of the performance measures of DT.

▪ Scope - DT has broader scope through out lifecycle and process whereas, traditional
simulations focuses on use case.

▪ Dynamic- DT is flexible, incorporate various models and matures with time thorough out
the asset’s life cycle. On the other hand, simulations are rigid in case of incorporation of
other applications and generation of new knowledge once a model is built.

▪ Complexity - Development of DT requires fusion of multidiscipline approaches and multi

technologies. Transformation of data various multiple sources and requirement of
extensive computational resource make the development of DT complex as compare to
conventional simulation.

9
Part II: DT types and classification.
 Digital Twin - Classification

11
 Classification based on functionality

▪ The plant twin: The plant twin is akin to 3D virtual reality

model. This allows for the stakeholders to run a series of
what if scenarios to plan the future activities in advance
and train personnel beforehand.

▪ Process twin: is digital replica of on plant processes.

This type of twin is used to study current process and
operation parameters to optimize process performance.

12
 Classification based on hierarchy
The difference among these groups is reflected in the level of
detail and accuracy requirement

▪ Equipment level: Equipment level twins are the most detailed

and accurate as they include detailed engineering drawings,
designs and data of the equipment being modelled

▪ System level: is an integrated version of multiple equipment

level twins. The high accuracy requirement is in general at the
aggregated level rather than of equipment it is built upon.

▪ Plant level: The multiple system level twin is further aggregated

to create the plant-level twin. The plant-level twin thrives to
imitate the overall plant behavior rather than system- or
equipment-level behavior.

13
 Classification based on Input Technology
▪ DT 1.0: incorporates operational technology (OT). Operational data from sensors
such as pressure, temperature, etc. are fed into models and algorithms for real-time
predictions.

▪ In DT 2.0: the information from OT is fused together with data from information
technology IT (CMMS, ERP etc.). This creates a much richer knowledge base for
decision making.

▪ DT 2.1: fusion of contextual information (geographical area, topography etc.) with IT

and OT. Contextual information may provide us answers to question about why the
two identical machines behave differently.

▪ DT 3.0: fuses Engineering Technology (ET) with IT and OT. Data from IT and OT and
contextual information is not enough to make reliable decision and predict future.
Subject engineering knowledge is needed along with data science.

▪ DT 4.0: shall incorporate artificial intelligence [AI] to learn from traditional black swan
events and able to predict the unseen event of similar scale and magnitude.

▪ DT 5.0: shall self-evaluate be able to recognize and adjust to the new normal.

14
 Classification based on modelling approaches

▪ Stochastic models: models the behavior of the entire

system / plant in terms of a probability distributions. For
example, data of multiple choke valves will be gathered
and what if scenarios will be analyzed in term of
probability distribution rather than single values

▪ Real time models: model’s behavior and states for each

element individually. For example, each valve will have
its own DT modelling its behavior instead of getting the
results in term of a probability distribution.

15
 Classification based on capability
Classified on a scale from 0-5 in terms of capability level

▪ 0-standalone: virtual 3D model disconnected from its physical asset

▪ 1-descriptive: connected to its physical counterpart via information

flow through sensors and SCADA systems

▪ 2-diagnostic: the capability to detect faults and abnormalities within

behavior and state of actual asset

▪ 3-predictive: able to make prediction about future production rate

and remaining useful life etc

▪ 4-prescriptive: prescribe the stakeholders “what to do” and “when”

▪ 5-autonomy: has the capability to automate tasks paving way for

unmanned facilities
16
Part III : Subsea Maintenance
 Subsea Maintenance
▪ Subsea systems are in general prone to wear and tear due to usage,
ageing and exposure to harsh environment.

▪ The degradation ultimately leads to system failures, which may result in

production and economic losses. In case of safety critical equipment,
this failure may lead to loss of human life and environment.

▪ The degradation can be slowed down or reversed by right maintenance

activities.

▪ However, the maintenance of subsea system is challenging. From

technical point of view lifting of whole subsea module, sending out
unmanned remote vehicles etc. are complex operations.

▪ Economically the cost of the maintenance interventions could be as

high as the cost of production loss due to degradation induced failures

18
 Maintenance Development in Subsea

find an optimum balance between the risk (production loss) and opportunity
(high reliability)
, the maintenance interventions need to be well planned in collaboration with
operations.

Proactive
Condition base
Time Based Maintenace
/Predictive
Preventive Maintenance
Maintenance
Run to failure
Corrective
Maintenance

Shift from traditional corrective, calendar - or age-based maintenance

polices towards condition-based maintenance.

19
 CBM and PHM
The central idea is that maintenance is only implemented when there
is enough evidence of degradation (imminent probability of failure of
equipment) derived from observed condition of the system.

20
 CBM and PHM
Diagnosis : “Where are we at this moment”

Fault detection: whether a fault is present or not.

Fault isolation: is about identification of the nature

of the fault.

Fault estimation: is about estimating the severity

of the fault. ¨

Prognosis : “where we will be in

future”
- Degradation Modelling
- Remaining useful life RUL

21
 AutoPro
PHM Process integration in DT approach

22
 Why Need for DT?
major technological advancement, social and political trends are shifting
with focus on green energy, low carbon solutions and digitalization

Profit margins are contracting , increase in market competition. Drop-in oil

prices

the harsh operating environment and the positioning of components on the

seabed make the accessibility, maintenance and inspection/monitoring of
subsea systems complex, costly and demanding.

This calls for digitalized method, remote operations, online

performance and condition monitoring.

23
Part IV : DT based PHM : Opportunites and
Challenges
 DT based PHM : Opportunities
Real time Monitoring and Control – DT allows for using real-time sensor
data from the subsea components allowing for real-time diagnostics and
prognostics resulting in capturing actual degradation for real time control
and decision making.

Maintenance Optimization - RUL estimations based on actual operating

conditions and parameters will help maintenance staff in formulating more
reliable maintenance, repair and replacement plans. Thus, to a large
extent eliminating unnecessary interventions and down time.

Synchronization of Operation and Maintenance - maintenance personnel

need to align the preventive tasks with the operational personnel
considering production plans, operational windows opportunity available
for maintenance interventions etc. Digital twin representation enables for
such effective synchronization of operation and maintenance

25
 DT based PHM : Opportunities

Remote Operations and Unmanned Facilities – Maintenance

personnel will be able to monitor the health of system
onshore, dispatching the staff when necessary. Reduction in
downtime and losses caused by human errors

Cost reduction and CO2 Emissions – PHM will realize

unmanned facilities and remote operations, this will in turn
result in cost reduction in terms of manhours requirements
and logistic, which will also result in lower CO2 emissions
and help to achieve the CO2 emissions targets.

26
 DT based PHM : Challenges
▪ Models and analytical algorithms capturing uncertainty
and impact of maintenance actions

- Diagnostic and prognostics models at system level

- Capturing Impact of maintenance action and uncertainty on degradation
process w.r.t. to future solicitation, environmental variations ,etc.

▪ Online algorithms

27
 DT based PHM : Challenges
▪ Multiple Twins synchronization

28
 DT based PHM : Challenges
▪ Sensor Drift – Digital twins gather data from various sensor
connected to its physical counterpart. The sensors are prone
to drift phenomena which affect the accuracy of sensor
readings. This challenges the reliability and credibility of the
study findings

▪ Fusion of Multidiscipline – DT based PHM which will require

scientist from various sciences especially production, data,
risk and reliability to join hands. Fusion of knowledge among
these disciplines will require to reach common goal.

▪ Validation - Models and simulations are to large extent in

general are based on assumed or randomly generated data.
Validation of the models in real settings is of great importance
for large scale implementation of DT based PHM.

29
Conclusion
▪ The difference is substantial and explicit in terms of real-time, scope, dynamics and
complexity

▪ The crucial difference being that digital twins have the power to utilize real-time state and
process unlike conventional simulation technology.

DT based PHM has its own characteristics which poses new challenges. Some of
which to be addressed are:

▪ Development of probabilistic digital twins that can capture uncertainty in component/system

degradation and remaining useful life with respect to future solicitations, environment
variations, degradation process and concurrent faults etc.

▪ Mathematical models and algorithms that can integrate multiple twins (for e.g., “maintenance
and production twin”) to optimize maintenance.

▪ Upgrading of the existing models or develop new algorithms that can be automated to run in
real-time.

▪ Validation of DT models in real settings and data.

• Thank You for Listening!

Questions and Comments?

31
 Bibliography
• Grieves M. “Digital twin: Manufacturing excellence through virtual
factory replication,” White paper, 2014.

• Altamiranda E, Colina E, "A System of Systems Digital Twin to

Support Lifetime Management and Life Extension of Subsea
Production Systems," OCEANS 2019 - Marseille, Marseille, France
2019:1-9

• Diego G. Luleå University of Technology. Digital twins powered by

IIoT and Industrial AI [youtube video].
https://www.youtube.com/watch?v=a5VSQakos5o&t=537s; 2020.

• Thumeera RW, Wroblewski L, Petersen KB, Gosine RG, Lesley AJ,

Oscar DS, et al. Digital Twin for the Oil and Gas Industry: Overview,
Research Trends, Opportunities, and Challenges. IEEE access
2020; 8:104175–97.

32